Have you looked at Digest::CRC? It has the CRC routine itself written in C, so if you can supply the data to it in large chunks (instead of reading byte by byte), it should be much faster than Perl alone.